Positive and Negative Symptoms in Psychosis

Description, Research, and Future Directions

Philip D. Harvey, Elaine Walker, Philip D. Harvey, Elaine E. Walker, Elaine Walker

• Publisher: Routledge
• Year: 2013
• Topic: Psychology
• eBook ISBN:
  9781134933457
• Subtopic: Clinical Psychology
• Index: Psychology

First published in 1987. This volume presents a collection of chapters on varied aspects of psychotic symptoms, largely within the context of positive versus negative symptoms. These chapters cover a broad range of aspects of these symptoms, such as longitudinal course, cognitive correlates, biochemical and structural correlates, conceptual issues, and research methods. The majority of these chapters were presented at the SUNY-Binghamton/Cornell University conference on schizophrenia that took place on October 17-19, 1985, in Ithaca, NY. That conference was designed to provide a forum for the dissemination of information on psychotic symptoms in general, with the overriding framework of positive versus negative symptoms.

RELATING COGNITIVE PROCESSES TO SYMPTOMS: A STRATEGY TO COUNTER METHODOLOGICAL DIFFICULTIES

Raymond A. Knight Ph.D.

There is a growing consensus that some form of an early information input dysfunction constitutes a central deficit in schizophrenia. A substantial body of evidence (cf. Nuechterlein & Dawson, 1984b, for a review) indicates that an input dysfunction might predate the onset of the disorder and be a consistent trait of schizophrenics in both their acute and remitted stages. For example, input dysfunctions have been found in nonpsychotic schizotypic individuals (Balogh & Merritt, 1985; Braff, 1981; Steronko & Woods, 1978), and in a significant subset of children at increased risk for adult schizophrenia (Asarnow, Steffy, MacCrimmon, & Cleghorn, 1978; Nuechterlein, 1983; but see also Harvey, Weintraub, & Neale, 1985). In addition, input dysfunctions have been identified in the first degree relatives of schizophrenics (DeAmicis & Cromwell, 1979; Spring, Levitt, Briggs, & Benet, 1983), further supporting the potential of such deficits as genetic vulnerability markers for the disorder. Finally, input dysfunctions also seem to have prognostic utility (Cancro, Sutton, Kerr, & Sugarman, 1971; Elliott & Knight, 1982; Knight, Elliott, Roff, & Watson, 1986), and may serve as important predictors of the course of the disorder.

The hypothesis that deficiencies in such elemental cognitive/perceptual processes may contribute substantially to thought disorder symptoms has remained a viable one, despite its reliance for support primarily on armchair theorizing. Surprisingly few studies report data on the relation between laboratory tasks purportedly measuring cognitive deficiencies and symptomatology (e.g., Cornblatt, Lenzenweger, Dworkin, & Erlenmeyer-Kimling, 1985; Green & E. Walker, 1984; Heilbrun, 1980; Knight et al., 1986; Mintz & Alpert, 1972; Neale, Oltmanns, & Harvey, 1985). Recently, several investigators (George & Neufeld, 1985; Knight et al., 1986; Neale et al., 1985) have challenged the complacent acceptance of this situation and have encouraged researchers to address the problem of the crucial interface. The major purpose of this chapter is to suggest a research strategy that is aimed at bridging this gap between deficient performance on perceptual/cognitive laboratory tasks and the manifest symptoms of schizophrenia.

Methodological Problems

There are, of course, numerous methodological problems that plague the investigators of schizophrenia. These have been discussed often (Blaney, 1978; Chapman & Chapman, 1973; Cromwell, 1972; Neale & Oltmanns, 1980; Salzinger, 1984), and I will not repeat them here. Rather, I will focus on those problems unique to the symptom and cognitive processes domains, and on the difficulties that arise when one attempts to integrate these domains, because the strategy that I will propose has been designed specifically to address these methodological problems.

The three areas of difficulty all involve aspects of construct validation, broadly defined. First, many of the important symptom scales that are supposedly designed to delimit cohesive behavioral domains are inadequately defined and poorly validated (Knight & Roff, 1985; Sommers, 1985). Second, the specification of the cognitive processes underlying schizophrenics’ performance deficiencies on laboratory tasks has its own set of methodological difficulties (Knight, 1984). Although I have previously proposed strategies to address these problems (Knight, 1984), my solutions, as we will see, create their own impediments to exploring hypothesized process-symptom links. Third, there are indications that we are not likely to find simple one-to-one mappings from process to symptom manifestation. Thus, we are faced with the prospect of both having to identify different deficient processes that lead to similar symptoms, and possibly having to isolate particular deficits with multiple symptom manifestations. In the next three sections I will illustrate each of these problems.

Assessing Symptoms

The current controversy over positive and negative symptoms in schizophrenia illustrates the perennial problems encountered by those who attempt to carve the behavioral episodes of patients’ lives into phenotypic symptom attributions, to cluster these symptom constructs into syndromes (Crow, 1980a; 1982; 1985; Strauss, Carpenter & Bartko, 1974) or to isolate the core symptoms of the disorder (e.g. Bleuler, 1911/1950). Difficulties abound at many levels of conceptualization--in the simple operationalization of symptom constructs (e.g., Knight & Roff, 1985; Sommers, 1985), in the determination of which symptoms define a syndrome (e.g., Andreasen, 1979b; Crow, 1985; Lewine, Fogg, & Meltzer, 1983), and in the specification of what relation syndromes should have with each other (Andreasen & Olsen, 1982; Bilder, Mukherjee, Rieder, & Pandurangi, 1985; Lewine et al., 1983; Pogue-Geile & Harrow, 1984; Rosen et al., 1984). These are, of course, all problems in defining the nomological network of the positive/negative symptom theory (cf. Cronbach & Meehl, 1955), and indicate the basic insufficiencies that must be overcome.

The extent of these problems can be illustrated by considering the lowest level of the nomological laws in the positive/negative symptom theories, the relating of observable properties or quantities to each other, that is, the creation of basic symptom constructs. Although the operationalization of symptom constructs is a low level of conceptualization, it actually involves a substantial amount of inference, because the schizophrenic “symptoms” in the theory are not simple, single behaviors, psychological responses, or other “signs” (Alpert, 1985), but are rather defined either by a multitude of criterion behaviors, which are often vaguely specified, or by dispositions that lack behavioral anchors. Moreover, when criterion behaviors are provided, there is frequently little empirical evidence that such behaviors differentially characterize the intended symptom as opposed to other symptoms. This lack of specificity has been a more serious problem in operationalizing negative symptoms, which are supposed to reflect the absence or quantitative deficiency of normal functioning, than it has been in describing positive symptoms, which hypothetically reflect the presence of deviant functioning.

The scales purportedly measuring flat or blunted affect, a widely accepted negative symptom, illustrate these problems. In a review of the empirical literature on four of the behavioral domains used by two major scales assessing affective flattening--Andreasen’s (1982b) Affective-Flattening Scale and Abrams and Taylor’s (1978) Scale for Emotional Blunting--Knight and Roff (1985) found that several of the items on these scales did not differentiate between flat affect and depression. In fact, some of the items purported to specify “flat affect” were actually more likely to select for depression. Thus, the clinical literature, which served as the source for items on both of these scales, was wrong in the specificity it attributed to certain behaviors hypothesized to assess flat affect. Knight and Roff (1985) also pointed out that many of the “behavioral” items on these two scales were so lacking in adequate specificity, both in terms of the particular behaviors described and the standards against which relative assessments were to be made, that they would force investigators at different sites to establish their own standards (as they did, Boeringa and Castellani [1982]). This would produce assessments that vary across research sites, and would possibly lead to inflated estimates of reliability as investigators at particular institutions settled on their own idiosyncratic, but easily agreed upon, manifestations of such constructs as “incongruous mood.”

Knight and Roff (1985) provided a partial solution to the flat affect specificity problem. They pointed out that flat affect can be discriminated from depression, if the behavioral indicators of affective expression are examined in the context of specific demanded and intended affects rather than simply in isolation. Moreover, they demonstrated that because depression is more episodic than flat affect, the reliability of discriminating judgments can be improved by the addition of longitudinal data. Unfortunately, their first suggestion requires a degree of control over the interviewer’s behavior and a structuring of the interview content that is difficult to implement in typical clinical settings. Their second suggestion, the use of longitudinal data, is a luxury not often attainable, except retrospectively in many assessment procedures. More importantly, their suggestions do not directly address the pressing, but elusive problem of establishing an adequate theoretical model of deviant affective expression to guide research.

As I have indicated above, the problems do not end at specifying adequate symptom constructs. There are theoretical and empirical controversies about what symptoms constitute these hypothesized positive and negative syndromes (e.g., Andreasen, 1982; Andreasen & Olsen, 1982; Crow, 1985; Lewine et al., 1983; Lewine & Sommers, 1985), about whether the positive-negative distinction might not indeed be an oversimplified dichotomy (e.g., Bilder et al., 1985; Gibbons, Lewine, Davis, Schooler, & Cole, 1985), and about how these syndromes should and do relate to each other (e.g., Andreasen & Olsen, 1982; Knight et al., 1986; Pogue-Geile & Harrow, 1984). Thus, not only are we lacking an adequate nomological network at the symptom level, but also at the levels of establishing syndromes and specifying how these syndromes relate to each other.

Specifying Cognitive Deficits

Research aimed at specifying schizophrenics’ cognitive deficits has been hampered by the finding that schizophrenics tend to be deficient on numerous, seemingly unrelated cognitive tasks (e.g., Killian, Holzman, Davis, & Gibbons, 1984). It became obvious to many that finding schizophrenics deficient on any single task provided little insight into their cognitive capabilities (Maher, 1974). A differential-deficit strategy was proposed to address this difficulty. It was argued that if two dependent measures were employed and schizophrenics were compared to controls on both tasks, a relatively greater deficiency on one task could be taken as evidence for a differential deficit. This strategy was, however, shown to have a serious flaw. Chapman and Chapman (1973;1978) argued convincingly that an artifactual differential deficit could be produced in any single experiment by an inappropriate selection of dependent measures. They reasoned that if one accepts the notion of a general cognitive deficit in schizophrenia, differential performance on two tasks might occur simply as a function of the unequal discriminating power of the tasks chosen. That is, more reliable, and therefore theoretically more discriminating, tasks might produce group differences that are absent when less discriminating tasks are employed. To remedy this situation, the Chapmans (1973, 1978) proposed that experimental tasks must be made equivalent in both difficulty and reliability.

Unfortunately, the Chapmans’ solution leaves several unresolved dilemmas, especially for those investigators interested in specifying the underlying cognitive processes responsible for schizophrenics’ performance deficiencies. First, as Blaney (1978) pointed out, their exclusive focus on reliability and discrimination totally neglected problems of construct validity. Ensuring that tasks were equally reliable and difficult did not guarantee that such tasks were measuring homogeneous constructs. In the likely event that a task tapped multiple processes, it would be impossible to determine which process was responsible for any differences that were obtained. Second, matching tasks on psychometric characteristics, particularly difficulty level, often has unacceptable consequences for the process-oriented researcher, because frequently it can only be achieved at the expense of confounding the hypothetical processes being compared (Knight, 1984). In many information processing and cognitive paradigms, stimulus and task difficulty provide an essential manipulation for assessing underlying hypothetical processes. To match on difficulty in many paradigms would mean to unmatch on process. Thus, to require matched tasks would eliminate many powerful analytic tools from the process-oriented researcher’s arsenal. Third, Nicewander and Price (1983) have recently demonstrated that there are several instances in which reliability and discriminatory power are independent. Their results seriously question the Chapmans’ basic assumption that matching on reliability guarantees equal discriminatory power. Finally, the matched-task solution provides no apparent mechanism for comparing laboratory tasks to symptoms. Simple correlations of performance on single tasks with symptoms might be confounded with general performance deficiencies. Subtracting performance measures derived from matched tasks (e.g., Harvey, this volume) introduces the additional difficulty of the notorious unreliability of difference scores. Indeed, the problem is more acute when matched tasks are employed, which are likely to be highly correlated. The reliability of difference scores decreases as the correlation between tasks increases (Magnusson, 1966). Thus, the difference scores of matched tasks are likely to have very low reliability.

Fortunately for the process-oriented researcher, who focuses on identifying the theoretical mechanisms underlying task performance, other strategies for addressing the general deficit and psychometric artifact problems are available. Cognitive psychology has matured sufficiently that not only does it provide paradigms that yield precise, predictable results, but also, many of its structures and processes can be measured by several paradigms. Thus, the process-oriented researcher can assess a particular process from several perspectives. Although the outcomes of these different measures do not provide exactly equivalent estimates of the same theoretical quantity, they are sufficiently precise to permit specific a priori predictions of the function patterns that would occur under conditions of adequate and inadequate performance. This makes available to the investigator several strategies that can finesse the general deficit dilemma (cf. Knight, 1984 for details). All of these strategies involve forming a network of specific, reliable, theoretically related tasks that have been explored in depth in the normal literature, and using this network to assess various levels and types of information processing in schizophrenics. Theoretical models can be used to predict what pattern of results should occur across these tasks according to various hypothesized deficits. It is the pattern of performance across tasks, and not the comparison of performance on any one task to that of controls, that is important. In this approach, each task is analogous to a consistency hurdle (Meehl, 1978), and a model must jump all the hurdles (i.e., accurately predict the pattern of performance) to be judged valid.

Although the process-oriented strategies clearly have their own set of problems (cf. Knight, 1984), the proof of their value lies in their ability to advance our knowledge about the nature of schizophrenics’ cognitive deficiencies. Some success in deficit specification can, indeed, be claimed (e.g., Knight, 1984; Knight, Elliott, & Freedman, 1985). They have, however, also created a new set of problems for those interested in relating these cognitive deficits to symptoms. The process-oriented strategies, like the information processing model that spawned them, are concerned with identifying the theoretical mechanisms that underlie task performance. Information processing paradigms have added a different perspective to construct validation, because they have introduced qualitatively different methods for determining constructs (Embretson, 1983). They focus more on task variability than on individual differences. Embretson (1983) has called this new kind of construct validation construct representation. She has contrasted it with nomothetic span, which corresponds more closely to Cronbach and Meehl’s (1955) examples of nomological network. Nomothetic span emphasizes the network of relationships of a construct to other constructs, and like the nomological network it focuses on individual differences. Consequently, it concentrates on an individual’s relative performance on specific tasks, whereas construct representation stresses the nature and structure of thought processes underlying performance. Such differences in procedures and goals make it problematic to integrate the results of these two approaches.

Although my process-oriented solutions to the general deficit problem are aimed at isolating the mechanisms that account for the differences between schizophrenics’ and controls’ cognitive performance, to achieve this goal the strategies place emphasis on the within-group patterns of performance across theoretically related tasks and not on the comparison of the level of performance to that of controls. Thus, like the construct representation approach, these strategies are in practice more concerned with task variability than subject variability. This, of course, creates problems for integrating the results of this research with the individual-difference oriented assessment of symptoms. For instance, a simple correlation of performance on an individual cognitive task with an assessment of a particular symptom would not differentiate the contribution to that symptom of general performance deficiencies from the contribution of the specific deficit purportedly being measured. This holds even when the process-oriented strategies had helped us to identify the importance of the specific deficit being assessed. This problem shares many important characteristics with the more general issue of integrating construct representation with nomothetic span, and must be addressed by any research strategy that claims to study the relation of cognitive processes and symptoms.

Symptom-Process Equivocality

Obviously, the task of relating cognitive process and symptom domains would be greatly facilitated if a simple one-to-one correspondence existed. If this were the case, we would probably be much further along in our understanding both of the nomothetic span of symptoms in schizophrenia and psychosis, and of the construct representation of the processes responsible for these symptoms. Not only does our lack of progress suggest that the relationship is complex, but there are also several converging lines of tentative evidence in the research on disordered thinking that confirm this suspicion. Many of the disordered thought symptom constructs have better behavioral anchors than the flat affect scales I reviewed earlier (e.g., Andreasen, 1979b; Johnston & Holzman, 1979; Oltmanns, Murphy, Berenbaum, & Dunlop, 1985). Here, however, as in the case of flat affect, there is little evidence that the presence of a symptom construct actually reflects the operation of the same underlying process or processes in all patients manifesting the symptom, and such symptoms do not correlate with alternative thought measures in consistent ways across patient groups (Harvey & Brault, 1986). At best, all we have is evidence for the internal consistency of some of the behaviors used to assess certain thought symptoms (e.g., Andreasen & Olsen, 1982).

When two disorders manifest the same symptom, it is often assumed that they share a common process. Thus, the manifestation of the same thought disorder symptom in schizophrenics and manics has been attributed to the psychotic state that they hypothetically share. An alternative explanation must, however, be seriously considered--that cognitive processes unique to each disorder result in similar symptoms. Such an alternative hypothesis, which I will refer to as symptom-process equivocality, would explain why, whereas cognitive deficit studies are beginning to isolate cognitive/perceptual processes that appear singularly schizophrenic (cf. Knight, 1984; Nuechterlein & Dawson, 1984b), many thought symptoms lack diagnostic specificity (e.g., Andreasen, 1979b; Oltmanns et al., 1985; Pope & Lipinski, 1978). Symptom-process equivocality could also account for particular patterns of results that we have encountered in two follow-up studies of schizophrenics we have completed during the last decade. Although these studies by themselves do not afford compelling proof of equivocality, and they were clearly not designed to elucidate the nature of the processing differences, they do provide, when combined with the other converging trends, a strong case for the importance of structuring our research designs so that they are sensitive to the possibility of symptom-process equivocality. Because this point is so critical, I will describe in some detail these patterns of results, which suggest that high scores on internally consistent scales of disordered thought may result from different underlying processes in broadly defined groups of schizophrenics.

In our thirty-year follow-up study of the child guidance records of preschizophrencis (cf. Knight & Roff, 1983), we found that the same internally consistent, thought disorder symptom scale was both predicted by and concurrently correlated with different variables at different times in the course of the disorder. For instance, as can be seen in the path analysi...

Table of contents

• Cover
• Halftitle
• Title
• Copyright
• Contents
• a) List of Contributors
• b) Editors Introduction
• 1) Relating Cognitive Processes to Symptoms: A Strategy to Counter Methodological Difficulties.
• 2) Validating and Conceptualizing Positive and Negative Symptoms.
• 3) A Twin Study Perspective on Positive and Negative Symptoms of Schizophrenia.
• 4) Laboratory Research: Its Relevance to Positive and Negative Symptoms.
• 5) Negative Symptoms in Schizophrenia: Longitudinal Characteristics and Etiological Hypotheses.
• 6) Positive and Negative Syndromes in Schizophrenia: Typological, Dimensional, and Pharmacological Validation.
• 7) Positive Thought Disorder in Schizophrenia: Its Importance, Its Longitudinal Course, and Impaired Perspective as a Contributing Factor.
• 8) Thought Disorder and Measured Features of Language Production in Schizophrenia.
• 9) Cerebral Structure and Symptomology.
• 10) Electrodermal Activity and Symptomology in Schizophrenia.
• 11) Genetics and the Phenomenology of Schizophrenia.
• 12) Afterword.
• 13) References
• c) Subject Index
• d) Name Index